Hybrid Security and Fault Detection in Short-Haul Fiber Links: Integrating Wave Equation Inversion with 5σ Backscatter Analysis

Modern data center and metropolitan area networks (MANs) utilize dense, short-haul fibersegments where traditional Optical Time-Domain Reflectometry (OTDR) is severelylimited by pulse overlap, inherent noise (CRN), and ambiguity from multiple reflections.This ambiguity renders conventional fault detection and, critically, security monitoring,unreliable. This work presents a novel, hybrid methodology to resolve these challenges,delivering high-fidelity, autonomous fiber monitoring.The core of the solution is atwo-branch system. The first branch utilizes a Wave Equation Inversion (WEI) kernel tomodel and invert the complex, overlapping backscatter signal, enabling the unambiguous,high-resolution mapping of the fiber’s true reflectivity profile, κ(z). The second branchintroduces a rigorous statistical α(z) detection scheme that applies a 5σ threshold to thelocalized backscatter decay parameter to identify sub-threshold attenuation eventscharacteristic of stealthy, non-reflective fiber taps.The entire architecture is integrated intoan autonomous system powered by Physics-Informed Neural Networks (PINNs) forreal-time data ingestion and processing, with results delivered to an LLM-based RAGsystem for diagnostic and prescriptive maintenance output. This approach transformsfiber monitoring from a signal processing problem into an interpretable, mathematically rigorous, and deployable security solution